Expired50% confidenceFactExact time
In GPT-5.6, the common effective context limit in actual conversations is approximately 258k, which is about 95% of 272k and is also the truly usable window size in Codex.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
7/19/2026
First Seen
Valid until: 10/17/2026(expired)
Sources
Related Claims
Unverified128K上下文窗口意味着模型可以在单次对话中处理约6.4万汉字，早期GPT-4仅支持8K上下文79% similarUnverifiedGPT-4's single output limit is approximately 4,096–8,192 tokens (roughly 2,000–4,000 Chinese characters), and the input context window has been expanded to 128K or even longer79% similarUnverifiedThe single-session context windows of the latest models such as Claude and GPT-4 have expanded to 128K or more tokens, but they still cannot retain memory across sessions74% similarUnverifiedIt is recommended to keep the GPT-5.6 context window within 272k as much as possible in order to balance cost and quality.73% similarUnverifiedGPT 5.6 crammed the chat function into the Codex workbench, turning it into a floating window.73% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/559862API
curl https://kongchang.com/api/v1/knowledge/claims/559862MCP
get_claim(id=559862)